Media Review 2020: Thoughts & Reaction
Some thoughts and reactions from last week's Football Index media review. Expectations, timing, communications and the next steps? What impressed and what does it mean for the platform moving forward?
First, the announcement itself. For a full breakdown of details, please visit the Football Index announcement page here. The first phase of changes will go live in early September 2020 ahead of the 20/21 season, with the second phase ahead of the 21/22 season.
Having been promised communication in July as to the next steps for the changes to the media monitor, there was some concern over the lack of communication and social media hype which usually precedes a big announcement. Nearing the end of the day on Friday 31st, preparing for the weekend, when all of a sudden the notifications pop up. Media Review 2020 with two phases!!! Here we go. Football Index have made it just in time!
So let's have a closer look.
You don't need to look far in the Football Index community to find criticism of the current media system. The regular gripes with the current media monitor are well known and broadly fall within the below
Narrow range of sources (overwhelmingly UK based) and dominance of certain tabloids (eg. The Express)
Poor player recognition in article headlines (only picks up full names in the headline)
The point scoring system & bias towards certain words/ clubs
Uncertainty over controversial topics and stories
Bias towards English players with no foreign media outlets included
The expectation of course is that these will be addressed and improved significantly. What worked for Football Index as a startup is no longer fit for purpose, and a much more robust system is required. Especially given there are times in the year when media is the only focus for weeks or even months on end. Not to mention, the Covid lockdown where Football Index had this as a fallback option and USP when so many rival bookmakers were scrambling to find any live sport for which they could accept bets.
So let's address the points above in turn
Sources & Article gathering
Two solutions were identified by Football Index. How articles are collected and the addition of a new source/ change of weighting for some existing sources.
Football Index will move away from RSS feeds and into a new unspecified system. Very little to say on this until it has been tested and analysed, but it's clear from my own work with RSS feeds that articles can indeed be missed or filtered unintentionally as well as reliability issues with feeds from various outlets. In short, Football Index claim this will mean a greater number of articles from Sky Sports & The Times which is interesting but difficult to gauge the impact it will have on players' media credentials. For example, whether the Sky Sports articles will include the broader Sky Sports news channels, such as those designated to Bundelisga, Ligue 1, LaLiga & Serie A?
Alongside this, the biggest change to the sources is the addition of The Athletic. A very positive move in the right direction towards more serious in-depth sports reporting and a welcome addition.
My concern with these changes to the sources are that they are relatively minor (tweaks?) which can feel a little disappointing given the potential for this review, and I have to admit to expecting more. The opportunity to rebuild the Media system and give it a complete overhaul is there and that would inevitably include a full review of the sources used, with the expectation that some would inevitably be removed and replaced. Whilst the current problems could still be addressed with the article scanning, and player scoring, having a solid base of outlets to begin with would offer reassurance that this is a full scale review, incorporating all aspects of the media monitor.
With the new lineup of outlets, there still exists 5 tabloids (including the Daily Mail) making up 25% of the current media sources. With some outlets publishing far less or where feeds are rarely picked up (EFL & The FA for example) this still leaves a significant weighting towards tabloid press. To achieve a better balance, Football Index could have potentially approached this by replacing the tabloids with football specific outlets (teamtalk.com for example or similar) or phase them out with the introduction of foreign outlets to replace them in the long term. The study below taken from today (04/08/2020) for Jadon Sancho's media scoring further highlights the weighting which tabloids have on the total articles included as well as points allocated.
Jadon Sancho Media Scoring for 04/08/2020. Sample taken up to 18:25.
Such small changes are also surprising given the results from the survey, although of course, the question and responses here relates to the media monitor overall as well as the press titles.
More feedback from the media survey here.
Further, this would have presented an ideal opportunity to make wholesale changes at a time when it is welcome and expected, as opposed to in the future, where further changes have the potential to disrupt trading and cause uncertainty.
Having said this, The Athletic is a fantastic addition and the greater presence of Sky Sports and The Times will hopefully provide for a more balanced football focused scoring system where players are rewarded based primarily on their performance. In addition, the advantage here will be the familiarity Football Index have with the existing sources and how they are now well positioned to build an advanced algorithm for player recognition and points allocation based on these.
2. Player Recognition
Player recognition is arguably the biggest problem with the current media monitor, and the announcement identifies two major improvements in player recognition.
Scanning not just the headlines, but the opening paragraph, sentences or synopsis for players names and aliases. This was a suggestion put forward by many in the survey so is a welcome addition and meets expectations
Improving player recognition by adding more names and aliases. This is the more interesting part and will be discussed further below.
Expanding on the second improvement above, this line caught my attention.
The list of aliases for footballers will be continually reviewed and updated and will be open to suggestions from customers (though suggestions will need to meet company policy standards such as avoiding overlap with similarly named players). This will provide a more dynamic and fair approach.
A system for us traders to put forward suggestions on names and aliases for players is a great idea. The current communication channels whilst adequate, simply don't allow for efficient processing of feedback, and a system, which in a basic sense would allow for traders to submit names and aliases associated to each player would be most welcome. This could also be expanded for example, to include a flagging system for articles which have either been missed, incorrectly scored or contain controversial topics to be reviewed. As well as a reporting system for any articles which have been missed.
A basic example of how this may work is below.
Example Player Name/ Alias Submission
Example Player Media Article Review
One of the key discoveries in my analysis of foreign media was how difficult it is to create a system which processes articles and attributes and scores them correctly. We can't expect Football Index to get this right all of the time, and encouraging input from traders is a great step forward and will hopefully be a catalyst for better engagement moving forward. Alongside this, it's also hoped that there will be involvement of traders (the trader panel?) in the testing of the new player identification and scoring system. Given how passionate and attentive the trader community is, it would make sense for Football Index to utilise this for the benefit of the platform and improving what is a very complex and sophisticated system.
3. Point Scoring
In brief, Football Index have decided to move away from the current generic algorithm and create their own tailor-made list of words and points scoring system, which will inevitably be an improvement. This includes
removing the positive (+1) point associated with the word 'United' and presumably other club references ('Nice' currently +3 points & 'Angers' -3 points)
Allocating additional weighting towards performance related words
Stronger negative scoring for articles in relation to negative sentiment
In addition, as with players names, the opening paragraph/ sentence/ synopsis will be scanned to evaluate overall sentiment. This should make for a more accurate scoring system overall but concerns remain over how this will work in practice, given how articles can be interpreted and points allocated. There will presumably need to be extensive monitoring when this is first implemented with options for corrections and a system in place for feedback.
The weighting towards performance related media is a positive step. On matchdays it would be logical that the best performing player should also feature heavily in the media and payouts should reflect this. From a trading perspective, this of course benefits those premium players who regularly win or challenge for performance and media dividends, increasing their chances of winning both.
4. Controversial Topics & Articles Relating To Discrimination
The communication on this was generally good, with positive stories to be retained and negative stories removed. The logic behind this decision is sound, but the test will be in how this is implemented and how any subjective articles are handled. Some examples below highlight how positive and negative sentiment can become confused and unclear and will require some continual reworking of the algorithm, ideally with trader input informing it as suggested above. This is where it would also be great to have the feedback system in place (flagging articles to be reviewed if the sentiment is not reflected in the points allocation).
The announcement finishes with plans for phase 2 of the Media Review where artificial intelligence and machine learning will be used to evaluate articles and where global media will be introduced.
5. Foreign Media
Most interesting for me, was the reference in the announcement to global media. Football Index specifically mentioned translation issues and localised content, suggesting the intention is to go down the route of translating foreign media outlets rather than choosing the more simple solution of English language European news outlets, such as the eight included in my study of foreign media here.
Including the above would certainly be an easier option and my concern is the accuracy of translations, especially given the challenges faced with judging sentiment and point scoring for current articles in English. Having said that though, it would make sense to include the most popular media outlets from the respective country (L'Equipe, Bilde etc).
Finally, the last line of the announcement. A teaser for a further announcement this week which will hopefully include developments to the Matching Engine/ Order Books.
We’re very excited about the future of our Media Rankings. Next week, we will announce our plans for the product and platform going forward - with exciting features and enhancements.
Expanding on some notes earlier about judging sentiment, the below articles are some examples of flaws in the current scoring system and how positive and negative sentiment can become confused.
Example 1 - High Scoring article
Headline for Victor Osimhen (foreign sample source used for indicative purposes only)
Total points under current system = 140 points
Keywords Breakdown: Victor (+3), United (+1), focused (+2)
Correct Sentiment: Neither positive or negative
Example 2 - Incorrect Player Attribution
3 Headlines in one and players receive all points generated
Total points under current system = 120 points
Keywords Breakdown: United (+1), interest (+1), join (+1), defender (+2)
Correct Sentiment: Neither positive or negative
Example 3 - Positive sentiment not picked up (negative scoring & incorrect player attribution?)
Positive headlines not rewarded under current points scoring system and points allocated evenly to both players
Total points under current system = -20 points
Keywords Breakdown: defeat (-2)
Correct Sentiment: Positive
Example 4 - Incorrect attribution / Invalid article / frivolous news
Total points under current system = 100 points
Keywords breakdown: stunning (+4)
Example 5 - Two players in headline
Both players mentioned receive equal points award
Total points under current system = 120 points
Keywords breakdown: winning (+4), snubbed (-2), awards (+3)
Correct Sentiment: Negative for Ronaldo, positive for Dybala
Example 6 - Article with negative sentiment incorrectly awarded positive points
Negative sentiment is not picked up in points allocation. Negative points are converted to positive and added?
Total points under current system = 60 points (actually -60 points)
Keywords breakdown: United (+1), leave (-1), admits (-1), defeat (-2),
Correct Sentiment: Negative